Connexion
AMD Ryzen 9 9950X3D CPU Review & Benchmarks vs. 9800X3D, 285K, 9950X, & More
We put the 9950X3D through numerous gaming and productivity benchmarks, efficiency tests, and more
The Highlights
- 9950X3D is a 16-core, 32-thread CPU with a 5.7 GHz max advertised boost clock and 128MB of L3 cache
- The 9950X is a better value for pure productivity and the 9800X3D is a better value for pure gaming
- The 9950X3D is a compelling CPU for both heavy production workloads and gaming
- Original MSRP: $700
- Release Date: March 12, 2025
Table of Contents
- AutoTOC
Intro
The quick version up front: The 9950X3D is comparable to the 9800X3D in most gaming scenarios, sometimes trading places; in production, it’s similar to the 9950X. The biggest change has been to the setup, which AMD says should now be simplified from prior dual-CCD parts with one faster CCD and one extra V-cache CCD. Historically, setting this up properly has made it necessary to isolate drives. If you were to install a 7600X and upgrade to a 7950X3D later, the easiest thing to do would be a clean Windows install (although there were ways to avoid this). That should be fixed now, but we’re still keeping all our drives isolated.
AMD is launching its R9 9950X3D CPU. This is a 16-core, 32-thread part with a listed MSRP of $700. The $600 MSRP 9900X3D will be launching alongside it, but wasn’t sampled, which is normally not a good sign.
Editor's note: This was originally published on March 11, 2025 as a video. This content has been adapted to written format for this article and is unchanged from the original publication.
Credits
Test Lead, Host, Writing
Steve Burke
Testing
Patrick Lathan
Mike Gaglione
Camera, Video Editing
Vitalii Makhnovets
Writing, Web Editing
Jimmy Thang
But as we all know, MSRP often doesn’t hold at launch as new silicon gets sold at higher prices. We just uploaded an entire video digging into that. In either case, for these CPUs, we definitely wouldn’t pay over MSRP since the 9950X (read our review) is available regularly for $545, with the 9800X3D (read our review) for gaming at $480 (which is MSRP) as we write this. That may change, of course.
Today, we’re reviewing the 9950X3D. It’s been a long review cycle the past 3 months, so we’re going to keep this one simple and focus on the numbers.
9950X3D Overview
Let’s get straight into it today. We’ll start with the specs.
The AMD 9950X3D is part of the Zen 5 architecture that launched with the 9700X (read our review) and other CPUs last year. The 9800X3D swooped-in a little later and cleaned-up what was a confusing and messy launch, largely making major moves for gaming CPUs and giving us something to be excited about as it’s a really good CPU.
The 9950X3D is a 16-core, 32-thread CPU with a 5.7 GHz max advertised boost, 4.3 GHz base clock, and L3 cache at 128 MB. TDP target is 170W.
For comparison, the normal 9950X also has a max advertised boost of 5.7 GHz, base of 4.3 GHz, and TDP of 170W. These are shared. The cache changes, at 64 MB of L3.
The 9950X3D has two CCDs, with one of the two CCDs bearing extra cache. This is stacked vertically. As we described in the 9800X3D review, the cache this time is flipped so that it’s closer to the substrate than the lid, pushing the cores closer to the IHS. In the 9800X3D review, we demonstrated how this helped significantly with cooling.
9950X3D Testing
We’re keeping it incredibly simple this time. As always, you can find our test bench information published here.
For gaming tests, we have all new data including the latest Windows updates and microcode for everything. That means we’ve refreshed the data set and wiped out what we had, so every CPU you that has been run was done in the last 3 days or so. We got the important ones in there. For production, we were able to salvage a lot of data since it’s the same.
We’ve been completely buried by one GPU after another in an onslaught of benchmarks and follow-ups the last few weeks, so for this one, we’re sticking to the basics.
Let’s just get into it.
Frequency Tests
Frequency - Blender All-Core
Frequency analysis is up first. We do this testing to ensure the CPUs are functioning as expected and to help explain the performance later.
First up is the Blender all-core workload. In this test, the 9950X3D had a frequency plot that started at about 5250 MHz, but settled closer to 5020 MHz to 5080 MHz during testing. This chart is intentionally zoomed-in to make it easier to see, so the scale purposefully does not start at 0.
For comparison, the 9950X non-3D (read our review) had higher peaks, but similar valleys. It ranged from 5010 MHz to 5080 MHz. In terms of average frequency over the course of the test, the 9950X3D averaged 5038 MHz all core to the 9950X’s 5036 MHz, but the X3D CPU did so with fewer peaks and more level frequencies in the middle of its range. We think this will be beneficial to it in gaming.
The 9800X3D 5220 MHz all-core, putting it well above both. This will help it in some specific workloads, but obviously the lower core count will set it back elsewhere.
Frequency - Cinebench 1C
The next chart is for frequency in a Cinebench single thread workload. This has the 9950X3D up in the range of 5650 to 5725 MHz, which hits AMD’s advertised frequency of 5.7 GHz. The 9950X holds its frequency steadier and with fewer dips between tile cycles, but is overall comparable.
The 9800X3D holds 5225 MHz throughout the test so it’s lower than both when in a single-threaded workload in this situation.
9950X3D Game Benchmarks
Baldur’s Gate 3
Baldur’s Gate 3 is up now. This one had the AMD R9 9950X3D at 155 FPS AVG, technically becoming a new chart topper. The 9800X3D was our chart topper last time and is now functionally tied with the 9950X3D as the best CPU on the chart.
The good news is that the 9950X3D doesn’t appear to be suffering from its dual-CCD approach, so parking is functioning properly and the CPU is not hamstrung by its extra threads.
The 9950X3D outranks the 7950X3D by similar margins as the 9800X3D did: It’s 23% higher average framerate, with lows comparable for the average. The 7950X3D (watch our review) outdid the 7950X, which we’re using old data for but should be no greater than 2-3% different based on our study of this test, by 29%. That’s with proper setup for the 7950X3D this time.
As compared to the 9950X at 101 FPS AVG, the 9950X3D outdid it by 54%. The 9950X is closer to the 7950X (watch our review) for performance, which makes sense. This game really benefits from the extra cache.
In fact, an easy example of this is the 5700X3D (read our review) vs. the 5600X3D (read our review): In some games, the 5600X3D outperforms the 5700X3D because of its higher clock rate. In this instance, the cache and core count was more beneficial than the frequency.
The 5800X3D (watch our review) remains an excellent CPU, up at 119 FPS AVG. The 9950X3D and 9800X3D outrank it by about 30%. As for Intel, it remains uncompetitive here. The 285K is getting crushed by two prior Intel generations for reasons discussed in that review, and that’s with new Windows updates.
Stellaris
Stellaris is one of our favorite CPU benchmarks because it looks at time rather than framerate, which is the most tangible to a user and the most directly influenced by the CPU. Players of 4X or other grand strategy games like Total War with the campaign map, Galactic Civilizations IV with turn pacing (and that’s a great game, if you haven’t played it), and Civilization would also see value here.
For Stellaris, the 9800X3D and 9950X3D both perform at the top of the chart. The 9800X3D outperformed the 9950X3D with a reduction in simulation time of 5%. That’s near error, but not quite. This seems to be a combination of a higher base clock and utilization.
The 9950X3D is definitely working as expected, though, because it’s outperforming the 9950X significantly. The simulation time requirement drops by almost 15%, from 32.3 seconds to 27.6 seconds.
This is the one game where Zen 5 in particular had stronger gains over Zen 4, with the 9700X doing well here and proving that. That’s from IPC uplift overall, where Zen 5 is benefitted.
Intel’s 285K is competitive with the 7800X3D and 9700X, at least. The 14900K (read our review) and 14700K (read our review) are within error of each other.
Dragon’s Dogma 2
In Dragon’s Dogma 2, the 9950X3D leads the chart. It landed at 132 FPS AVG here, passing the 9800X3D by a measurable but irrelevant 3.2%. Both CPUs lead all of Intel’s, although Intel at least lands its prior two generations ahead of the 7950X3D and 7800X3D with the game’s updates. This game really seems to benefit from extra cache, with the 9950X3D leading the 9950X by 46% and the 9800X3D leading the 9700X (although they have other differences) by 41%. Dragon’s Dogma 2 remains heavy on CPUs in NPC-intensive areas.
The 285K continues to impress with how much of a downgrade it is from not only AMD’s current generation, but Intel’s past generations.
We added the older results for the 3700X and 2600 to this chart. We noticed that performance on older generations hasn’t changed much. At most, there might be a 5% change here, but we don’t think so. Even with that though, anything is an upgrade.
Intel has seen the most upgrade since our last round of tests. This game has gotten updates, so it’s possible some of those were targeted at Intel. Windows updates could also affect it. We consistently saw uplift across Intel’s CPUs. That’s shifted the relative ranking of the 14th and 13th gen against the 7800X3D (watch our review).
Final Fantasy XIV Dawntrail - 1080p
Final Fantasy 14: Dawntrail is up now. In this one, the 9800X3D ran at 380 FPS AVG, with the 9950X3D at 373 FPS AVG. We observed relatively wide run-to-run variance in some of these results, so the error bars are wider than typical. The 9800X3D leads the 9950X3D by just 2%, so they are functionally equal.
The 9950X3D bests its 9950X non-3D variant by 50 FPS or so, or 16% here in average framerate. The 1% lows are also significantly improved, indicating that frametime pacing is keeping up with improvements in the average framerate.
The improvement over the 7950X3D is 5.8%.
Intel’s closest CPUs don’t appear until the 14900K at 310 FPS AVG. This is mostly interesting because there was a time when Final Fantasy’s prior benchmark versions were entirely dominated by Intel, with the clean division halfway down the chart. This is actually what we’re seeing now favoring AMD, relegating Intel to the bottom. That’s flipped in recent years and generations.
Intel’s 285K underperforms against its prior two generations. There was no change in performance against last time for the 285K. Intel’s one advantage in this test is frametime pacing, where the 0.1% lows indicate that Intel’s CPUs generally have more consistent frame-to-frame intervals than AMD’s CPUs, although not by an amount that’d change your experience in a noticeable way.
The 5600X3D outperforms the 5700X3D in this game. This has been known and is because of the higher clock speed on the 5600X3D, which proves more valuable than the extra cores.
Final Fantasy XIV - 1440p
At 1440p, the top of the chart truncates as a result of GPU limitations on the RTX 4090 (watch our review). We’ll move to a 5090 (read our review) for our full revamp of CPU testing for the next major architecture, but for now, this is where we cap-out. We’re sure this is deeply disappointing to all 12 of you who have an RTX 5090.
The 9800X3D and 9950X3D are about the same once again. The 7950X3D is also now about the same, as is the 9700X, thanks to the external limitations. This is a good reminder that the gains once scaling graphics are most seen in time-based situations or in seriously heavy CPU games like Dragon’s Dogma 2, but otherwise, most of the time you’ll get the most uplift from a GPU.
Starfield
In Starfield, we had the 9950X3D at 171 FPS AVG, leading the 9800X3D’s 165 FPS by 3%. The 9800X3D was notably ahead of the 7800X3D and the 9950X3D continued that, though neither had as revolutionary of a gain as we’ve seen in other benchmarks.
The 14900K trails the 7800X3D, improving upon its prior round result in a meaningful way; however, because of the improvements we’re seeing in the prior generations, the 285K now falls back behind Intel’s 14900K in this test. The 285K still regresses and generally embarrasses Intel, trailing even the 13700K (watch our review). Intel has continually tweaked its microcode on these prior generations, so it’s possible that they rolled-out a microcode that had lost some performance at some point and they’ve regained some now with the 13th and 14th series. We update BIOS to the newest version for each round.
Against the 9950X at 124 FPS AVG, the 9950X3D improves by 37%. That’s one of the larger gains. Of course, if you’re not going to use the extra cores, the 9800X3D makes more sense for value.
Cyberpunk 2077 - 1080p/Medium
Cyberpunk 2077 is back in our CPU test suite again with the Phantom Liberty expansion. Tested at 1080p/medium here, the 9800X3D and 9950X3D both ran at about 219 FPS AVG and were well within run-to-run variance at only fractions of a frame per second apart. The 7800X3D trails, but not by much. It’d be roughly the same experience as these two.
The lead of the 9950X3D over the 9950X is 37% again, matching some of the other games. The Intel 200 series outdoes the prior generations here, finally, with the 285K at 170 FPS AVG. Unfortunately, that’s still below the AM4 5700X3D and 5600X3D.
F1 24 - 1080p
F1 24 at 1080p is up now. This one has the 9950X3D and 9800X3D again roughly tied, with the 7800X3D not far behind. The advantage is only 7%. The 9950X3D leads the 9950X non-3D variant by 29%, slightly reduced from the advantage seen in other games. We might be hitting a GPU limit here.
Intel’s 14900K is its closest competition, released in 2023, with the 285K down at 9950X levels. The 5600X3D and 5700X3D results show again that this game likes frequency and IPC to some extent.
F1 24 - 1440p
1440p is almost exactly the same in the bottom half, with the top switching around due to GPU overhead and limitations on GPU scaling. The 5800X3D falls down the ranks as the 14th and 13th gen handle the overhead a little better and with more stable frametime pacing, which helps the average. Otherwise, things are about the same sans limitations of scaling for the 9950X3D.
9950X3D Production Benchmarks
Grab a GN15 Large Anti-Static Modmat to celebrate our 15th Anniversary and for a high-quality PC building work surface. The Modmat features useful PC building diagrams and is anti-static conductive. Purchases directly fund our work! (or consider a direct donation or a Patreon contribution!)We’re moving on to production tests now. This is where the 16-core CPUs do well. Historically, we’ve seen the X3D variants of all of these CPUs underperform against the non-X3D parts due to power allocation to allow higher boosting. Extra cache doesn’t help in our testing here normally. We’ll see if the 9950X3D breaks that general trend.
Blender Rendering
Blender testing hasn’t changed since our October round. We ran validation on several CPUs and the results came out basically identically, so we can keep a lot of data for more comparisons. This should help those of you on older hardware because we’ve got more present here.
The 9950X3D required 6.6 minutes to complete the render, which is about tied with the 9950X. It was technically faster, but in reality, these are tied. That’s good news for the X3D part, though: Past X3D CPUs, like the 7950X3D, have been technically slightly slower than their non-3D equivalents. That’s not because of scheduling or parking, but because the frequency is slower in an all-core workload.
Another good example is the 7800X3D, which was slower than the 7700X by time required, or 5800X3D as slower than the 5800X. The 9950X3D is the first to break this trend in a big way. Technically, the 9800X3D looked like it was doing that against the 9700X, but the power target was what brought most of that change.
The 9950X3D outperforms the 285K here, with about a 7% reduction in total time required to complete the render.
Chromium Code Compile
Chromium code compile in Windows is another where the data set hasn’t changed, so we were able to salvage it after validation. The 9950X3D required 81 minutes to complete the compile, which is comparable to the time required for the 9950X, but technically improved. This puts it marginally ahead of the freshly retested 285K, with a reduction in time required to compile from 285K to 9950X3D of 4.7% less time required. The 14900K required 88 minutes here, with the 265K at 98 minutes.
The 9950X3D is the new leader in our compile test. This is not going to be representative of every type of code compile, just like none of these tests is representative of every angle of a use case; however, the way we test it, the 9950X3D is the new leader short of going to Threadripper.
7-Zip Compression
Data for decompression and compression can’t be salvaged, so it’s all new.
In 7-Zip file compression testing, the 9950X3D led the chart at 206,643 MIPs, or millions of instructions per second. That has it just ahead of the 9950X by 3.3%. This is one of the tests where cache can help, depending on its implementation. The 5600X3D and 5600X are good examples of this: The X3D part has a lower advertised frequency, but manages to still roughly tie the 5600X.
The 9950X3D outperforms the 7950X3D by 8.5%, which completed 191K MIPS. The 14900K is next at roughly 189K MIPS, then the 13900K (watch our review). The 285K follows all of these, down at 179K MIPS.
Core count clearly matters in this test: The 3950X 16-core CPU is outperforming the 5900X 12-core CPU and 9700X 8-core CPU.
7-Zip Decompression
In 7-Zip Decompression, we measured the 9950X3D at 277K MIPS, with the 9950X non-3D at 272K MIPS. You wouldn’t really benefit from the 9950X3D in a meaningful way in either compression or decompression in this workload. The 9950X achieves all of the performance already, so you’d need use cases that more directly leverage the cache to get value out of the 9950X3D.
Intel’s 14900K is its closest competitor, followed by the 14700K and then the 285K.
Adobe Premiere
We saved some of the data for Adobe Premiere as well. The biggest swing was to Intel’s 12th to 14th Gen CPUs here, where we saw some movement from the Windows updates recently. Most of the other parts stayed relatively stationary. Any 12th to 14th Gen CPUs with data prior to this round would move around a bit, so be aware of that; however, just to try and offer some extra data that’s still mostly comparable, we’ve left those parts here. Most of this data is brand new.
The 9950X3D scored 11600 points in the Puget suite aggregate extended scoring for Premiere, which puts it at the top of the chart. It bests the 9950X by 5.8%, with the 14900K closest to it, then the 285K. The improvement over the 7950X non-3D is 7%.
9950X3D Efficiency
We’ll keep power and efficiency testing short this time and just show a couple situations.
Starfield
In Starfield, the 9950X3D ended up at 1.7 FPS/W, putting it behind the 7950X3D and 7800X3D, but tied with the 5700X3D and 9800X3D. The 9950X3D pulled 98.8W when playing this game, and Starfield is one of our games that most heavily loads the CPU (but is still nothing like an all-core Blender workload).
The 9950X non-3D part pulled 168W in this same test, putting it down at 0.7 FPS/W. That means the 9950X pulled nearly 70W more than the 9950X3D, or about a 70% increase in power consumption despite running at a lower framerate. In terms of FPS/W, the 9950X3D is both higher framerate and lower power, and so it is far more efficient. It’s still not as efficient as the low-power 7800X3D, though.
7-Zip Compression
7-Zip compression shows that the 9950X3D can still be power-hungry. In our compression efficiency testing, the 9950X3D pulled 203.8W. That put it at 1014 MIPS/W, which makes it less efficient than about half the chart. The CPU is the best performer, but not for efficiency and that’s because it’s pulling 204W, its efficiency has decreased compared to some others.
The 9950X scored 979 MIPS/W and pulled the same power at 204W, making it less efficient than the 9950X3D. The 7800X3D is a lower performer overall, and in big ways, but has such impressively low power consumption that it ends up being the most efficient.
Of course, if you were serious about running this kind of workload all the time, you’d still want something more powerful than the 7800X3D.
7-Zip Decompression
Decompression testing looks better for the 16-core parts, with the 7950X3D proving incredibly efficient here, followed by an impressive result from the 7950X non-3D with ECO Mode enabled. The 9950X3D ran at 1358 MIPS/W, putting it slightly ahead of the 9950X. They’re still in the middle of this chart though.
9950X3D Conclusion
Visit our Patreon page to contribute a few dollars toward this website's operation (or consider a direct donation or buying something from our GN Store!) Additionally, when you purchase through links to retailers on our site, we may earn a small affiliate commission.That’s it. You have the numbers.
For the quickest recap: For gaming, you can think of the 9950X3D like a 9800X3D. We didn’t run into any major issues with the 9950X3D here. That in and of itself is kind of an accomplishment for AMD. The company has really struggled over the years with the dual CCDs, where one has the extra X3D cache on it. Over the years, it’s taken them some time to get to a place where it’s not regressive and where it’s a little easier to set up. The 9950X3D does appear to do that in our experience so far and that is a major improvement for AMD. It’s taken them some generations to get there.
If you have the funds and are looking to build a purely gaming computer, we think you should scale it down and go for a 9800X3D. It’s just not that big of a difference as the 9800X3D often trades places with the 9950X3D and you save some money.
Intel, on the other hand, is out of this conversation right now. They are not part of the high-end expensive CPU for gaming build scenario at the moment.
Meanwhile, the 9950X makes sense for production-heavy builds that don’t have an explicit use for that extra cache. There’s definitely use-cases for this out there. We see a little bit of that in our 7-Zip testing, but for the most part in the things we test, it doesn’t tend to benefit from the extra cache in non-gaming scenarios.
Where the 9950X3D, and the other X3D 16-core parts, shine is a more limited use case where you have some mix of really heavy production and really heavy gaming. If you do a lot of compression, decompression, maybe render things on the CPU, are heavy into Premiere, or do a lot of code compiles and play a lot of games, then that’s kind of the use case for the CPU.
If you’re in one camp or the other exclusively, then you can save some money by going for either a 9800X3D or a 9950X.
We wouldn’t pay more than MSRP for the 9950X3D. CPUs tend to stick closer to MSRP, but can still have stupid prices from some retailers or third-party sellers.
.png){kind=link}